# AI-Assisted Operations Request Triage Workflow

## Problem
Workplace requests can arrive inconsistently and require manual interpretation, prioritisation, information extraction, response drafting and follow-up.

## Proposed solution
Incoming request → AI classification → structured extraction → missing-information check → draft response → human review → final action.

## Dataset
20 synthetic company-style requests created specifically for this project. No real confidential information was used.

## Test coverage
Includes clear requests, incomplete requests, ambiguous urgency, conflicting records, technical access problems, financial approvals, credential handling and explicit deadlines.

## Evaluation
AI outputs are compared with manually defined expected classifications. Record category accuracy, urgency accuracy, missing-information detection, risk detection and appropriate human escalation.

## Limitation
The project demonstrates workflow design and evaluation; it does not represent deployment inside a real organisation.